Clustering in computer systems means the clustering of multiple computer systems together with communication links between them. As the computing system needs more room to grow, another computer or group of computers may be added to the system thus increasing both storage and processing capabilities of the computing system cluster.
In general, clustering systems connect multiple computers, to function together as a single unit. Administrators of a cluster use an interface to monitor and manage the cluster. One common technique is a browser-based management system. For example, the administrator runs a browser that connects via a network to a web server on one of the cluster nodes. The web server receives requests, gathers the necessary data, and sends the results back to the administrator.
The administrator can perform a variety of monitoring tasks to determine the status of the cluster. For instance, the administrator may wish to see if any cluster nodes have failed, if there are any hardware or software problems, if there are any changes to the configuration, or if the status of any of the hardware or software components has changed.
One difficulty of monitoring cluster configuration through a browser is that web pages are typically static. Once the page has been loaded from the server on the cluster, the page will remain fixed, even if the cluster configuration or status changes.
A simple solution to this problem is to periodically automatically reload the web page, e.g., every 30 seconds. The disadvantages of this method is that the web server must constantly keep providing pages, even if nothing has changed, and the displayed state is not quite up-to-date as changes are delayed by up to 30 seconds. Reloading more often worsens network traffic issues as well as degrades browser performance (e.g., constant reloads will reset the page thereby making reading of the page difficult), while reloading less often worsens the problem of stale information being provided by the browser.
A second solution keeps a connection open between the browser and the server and provides new pages whenever a change occurs, using a technology such as HTTP Server Push. This technique puts a considerable load on the server, since the server must keep track of each browser accessing the data, keep a connection open to each browser, and handle all updates.
Another difficulty of monitoring cluster configuration is that the browser is connected to a single node of the cluster, while changes in configuration or status can occur on any node.